Crash prediction models are extensively used in highway safety analysis. In this paper, we introduce a recently developed data mining technique to predicting motor vehicle crashes, that is, the multivariate adaptive regression splines (MARS) technique. MARS has promising prediction power, and does not suffer from a “black-box” limitation. Negative Binomial (NB) and MARS models were fitted and compared using extensive data collected on unsignalized intersections in Florida. Two models were estimated for rear-end crash frequency at 3 and 4-legged unsignalized intersections. Treating crash frequency as a continuous response variable for fitting a MARS model was also examined by normalizing crash frequency by the natural logarithm ofthe annual average daily traffic. Finally, combining MARS with a machine learning technique (random forest) was explored and discussed. The significant factors affecting rear-end crashes were traffic volume on the major road, the upstream and downstream distances to the nearest signalized intersection, median type on the major approach, land use at the intersection’sinfluence area, and the geographic location within the state. The study showed that MARS can predict crashes almost similar to the traditional NB models, and its goodness-of-fit performance is quite encouraging. Also, using MARS for predicting continuous response variables yielded more favorable results than predicting discrete response variables. The generated MARS models showed the most promising results after screening the covariates using random forest.
Abstract